Target off-grid parameter estimation method for MIMO-OTFS radar communication integrated system

Abstract

The invention relates to a target off-grid parameter estimation method for an MIMO-OTFS radar communication integrated system, and belongs to the field of radar communication integration. Comprising the following steps: constructing a receiving signal model of the MIMO-OTFS radar communication integrated system; transforming the received signal matrix, and expressing a target parameter estimation problem as a sparse signal recovery problem; carrying out rough estimation on the target parameters by adopting a 3D-OMP algorithm based on a grid; establishing an off-grid initial search interval according to the parameter rough estimation value; and realizing accurate estimation of target distance, speed and azimuth angle parameters from the grid by combining a three-dimensional golden section thought. The method has the advantages that an off-grid 3D-OGOMP algorithm is provided, the problem of grid mismatch existing in a traditional compressed sensing algorithm can be solved, and the operation efficiency can be improved by combining three-dimensional golden section search. When the method is applied to an MIMO-OTFS radar communication integrated system, high-precision and low-complexity target parameter estimation can be realized, and reference is provided for intelligent traffic application.

Description

Technical Field

[0001] The present invention belongs to the field of radar communication integration, and relates to a target parameter estimation technology of an OTFS-MIMO radar communication integration system, and in particular to a target parameter estimation technology of a MIMO-OTFS radar communication integration system based on off-grid compressed sensing. Background Art

[0002] With the large-scale deployment of the fifth generation (5G) mobile communication system around the world, the number of wireless devices has exploded, resulting in the breakdown of the harmonious coexistence of communication and radar perception, and the emergence of many disadvantages such as electromagnetic spectrum congestion and interference. In order to alleviate the pressure on the existing frequency bands, and considering that radar and communication systems have many similarities in physical concepts and hardware implementation, in recent years, researchers have proposed the concept of radar communication integration (JRC), which shares frequency bands and hardware platforms between communication and radar systems, aiming to use the same hardware platform and signal waveform to simultaneously realize the functions of communication and radar perception. JRC has become an important development direction of the future sixth generation (6G) mobile communication technology. Among them, radar perception based on communication signals is a key technology of the JRC system. Especially in intelligent transportation systems (ITS), achieving accurate perception of target vehicles has great application potential. How to explore high-precision, low-complexity joint estimation methods for target distance, speed, angle and other parameters is an important issue worthy of in-depth study.

[0003] In the early stage of the research on target parameter estimation in the JRC system, the most commonly used integrated waveform was the orthogonal frequency division multiplexing (OFDM) signal due to its high spectral efficiency and strong target detection robustness. However, OFDM modulation technology is highly sensitive to Doppler frequency shift. In high-speed mobile scenarios and millimeter wave (mmWave) frequency bands, the inter-carrier interference generated will destroy the orthogonality of subcarriers, reduce the accuracy of target parameter estimation, and cause a significant deterioration in the performance of OFDM. In recent years, orthogonal time-frequency-space (OTFS) modulation has begun to emerge as a new communication modulation method. Unlike OFDM, which modulates communication symbols in the time-frequency (TF) domain, OTFS uses the delay-Doppler (DD) domain channel characteristics and symbol multiplexing. In OTFS, communication symbols are placed on a two-dimensional plane in the DD domain, which can resist the subcarrier orthogonality problem and has significant robustness to dynamic channels with Doppler frequency shift. Therefore, OTFS has become a new integrated waveform that is more suitable for sensing target distance and velocity parameters in the JRC system. In particular, MIMO-OTFS, which is formed by combining OTFS modulation with MIMO technology, makes the JRC system have higher spectrum efficiency and better signal transmission quality, which can not only estimate the target distance and speed, but also estimate the target azimuth.

[0004] In order to achieve fast and accurate target perception in the JRC system, it is crucial to develop a high-precision target parameter estimation method. At present, some OTFS-based target distance and speed joint estimation methods have been proposed, such as the maximum likelihood estimation (MLE) method. However, this type of method has a high complexity in distance and speed estimation in practical applications. In order to improve the system performance, some scholars have designed a method based on compressed sensing by utilizing the inherent sparse characteristics of the millimeter wave channel, and applied the orthogonal matching pursuit (OMP) algorithm to the target parameter estimation of the JRC system. However, the idea of ​​this method is based on grid division, and its defect is that the estimation accuracy is reduced due to grid mismatch, and high-precision parameter estimation cannot be achieved. Therefore, some scholars have further proposed a JRC system target parameter estimation method based on off-grid compressed sensing. The idea is: based on the rough estimation result based on the grid, the grid is refined again, thereby improving the accuracy of parameter estimation. However, the above method is limited to two-dimensional parameter estimation of distance and speed. For the MIMO-OTFS radar communication integrated system, how to realize the joint estimation of target distance, speed and azimuth angle three-dimensional parameters, improve the estimation accuracy and reduce the calculation complexity is a difficult problem that needs to be solved urgently in the target perception technology of the JRC system. Summary of the invention

[0005] The present invention provides a method for estimating off-grid parameters of a MIMO-OTFS radar communication integrated system. In view of the shortcomings of traditional estimation algorithms, a joint estimation method for target distance, speed and azimuth based on off-grid three-dimensional orthogonal matching pursuit (3D-OMP) is designed, which is referred to as the 3D-OGOMP method. In the parameter estimation stage, it is divided into two steps: 1) using the 3D-OMP algorithm to obtain a rough estimation result based on the grid; 2) combining the three-dimensional golden section method to realize off-grid parameter estimation, thereby obtaining accurate estimation results of target distance, speed and azimuth. Compared with traditional methods, the accuracy of target parameter estimation is greatly improved, which can be applied to intelligent transportation systems to achieve accurate perception of target vehicles.

[0006] The technical solution adopted by the present invention comprises the following steps:

[0007] Step 1: construct a receiving signal model of the orthogonal time-frequency-space (OTFS) modulation multiple-input multiple-output (MIMO) radar communication integrated JRC system;

[0008] Step 2, vectorize and sparsely represent the received signal matrix, and formulate the target parameter estimation problem as a sparse signal recovery problem;

[0009] Step 3: Based on the sparse signal recovery problem in step 2, a grid-based three-dimensional orthogonal matching pursuit (3D-OMP) algorithm is used to roughly estimate the target parameters.

[0010] Step 4, establishing smaller off-grid search intervals for multiple targets based on rough estimates of target parameters;

[0011] Step 5, combining the three-dimensional golden section idea, uniformly sample the initial off-grid search interval eight times, calculate the function value, update the search interval according to the maximum value, and after multiple repeated iterations, achieve accurate estimation of the target distance-speed-azimuth.

[0012] In the step 1, the present invention assumes that at the base station side, a classic OTFS resource unit structure is composed of M subcarriers and N symbols. In OTFS modulation, the orthogonal amplitude modulation QAM information symbol is regarded as being placed in the DD domain plane, which is recorded as:

[0013]

[0014] Where △f and T represent the subcarrier spacing and symbol duration respectively, and satisfy △f = 1 / T, the DD domain signal can be expressed in matrix form as Consider a system with N t transmit antennas and N r MIMO-OTFS radar communication integrated system with receiving antennas, using precoding matrix After spatial multiplexing, the DD domain signal transmitted on the i-th antenna is:

[0015]

[0016] Where W represents the Hadamard product operator. According to the need of target parameter estimation, the precoding matrix is ​​designed using the omnidirectional transmission strategy. Divide the entire data frame into mutually exclusive and exhaustive subsets to achieve N t The orthogonality between the antennas is determined by the symplectic inverse Fourier transform (ISFFT). The time-frequency TF domain signal transmitted on the i-th antenna is:

[0017]

[0018] in and are the M×M dimensional discrete Fourier transform DFT matrix and the N×N dimensional inverse discrete Fourier transform IDFT matrix, respectively. H represents the conjugate transpose operator, and the Heisenberg transform is performed on the TF domain signal to obtain the time domain signal transmitted on the i-th antenna:

[0019]

[0020] Among them, g tx(t) is the waveform function of the transmitted signal. Finally, a cyclic prefix CP is inserted into the time domain signal of each data frame. The time domain signal can be expressed as:

[0021]

[0022] Where T cp represents the cyclic prefix duration, That is the transmission signal of the MIMO-OTFS radar communication integrated system;

[0023] Assuming that there are P targets in the line-of-sight path LoS of the perception environment, the MIMO channel including delay, Doppler frequency shift and azimuth parameters can be expressed as:

[0024]

[0025] where α p is the channel gain, and denote the transmitting direction vector and the receiving direction vector respectively, (·) T represents the transpose operator, t p , ν p ,φ p and θ p They represent the time delay, Doppler frequency shift, DOD, and DOA parameter information of the pth target respectively. δ(·) represents the delta function. After transmission through the channel, the received time domain continuous signal is:

[0026]

[0027] in Indicated by The signal vector without CP is formed. Indicated by The signal vector with CP is formed. represents the additive white Gaussian noise AWGN vector that is independent of the signal;

[0028] exist m=0,1,…,MN-1 for the received time domain signal vector After sampling, the matrix form of the received signal is:

[0029]

[0030] in Indicates rounding up, Π MN is a cyclic shift matrix of dimension MN×MN, I N represents the identity matrix of dimension N×N, Indicated by The transmit signal matrix is ​​composed of Represents the sampled additive white Gaussian noise AWGN matrix. Considering a single-base radar system, the magnitude of DOD and DOA are the same, so both are represented by the symbol θ p represents the azimuth, is a diagonal matrix describing the Doppler frequency shift characteristics of the target, expressed as:

[0031]

[0032] represents the diagonal matrix containing the target delay effect, expressed as:

[0033]

[0034] These two diagonal matrices retain the real delay and Doppler frequency shift effects in the channel matrix and reconstruct the real physical channel. That is the receiving signal matrix of the MIMO-OTFS radar communication integrated system.

[0035] In the step 2 of the present invention, first, in order to use the compressed sensing theory to realize the target parameter estimation, the received signal matrix Y is vectorized and expressed as:

[0036]

[0037] Among them, vec(·) represents the vectorization operator. Since the number of targets in the entire space is small, the following sparse signal representation model can be constructed:

[0038] y=Φα+z

[0039] in, is a dictionary matrix, is a sparse vector, is the noise vector, G t , G ν and G θ are the number of grid points corresponding to the three parameters of delay, Doppler frequency shift and azimuth respectively. The purpose of the above sparse signal recovery problem is to recover the sparse vector α by receiving the signal vector y and the dictionary matrix Φ. Here, the vector of the jth column of the dictionary matrix is ​​expressed as:

[0040]

[0041] where j∈[1,G t G ν G θ ],j=mns,m=0,…,G t -1,n=0,…,G υ -1,s=0,…,G θ-1, therefore, if the values ​​of m, n and s are obtained, the formula and Calculate the corresponding time delay, Doppler frequency shift and azimuth values ​​for each column respectively.

[0042] In step 3, the present invention uses the 3D-OMP algorithm to obtain a rough estimate of the target delay, Doppler frequency shift and azimuth based on the sparse signal recovery problem in step 2. The specific process of the 3D-OMP algorithm is as follows:

[0043] 1) Constructing an overcomplete dictionary matrix Φ: The idea of ​​three-dimensional sparse signal recovery can be understood as searching on a three-dimensional grid set Γ, which can be expressed as:

[0044]

[0045] The vector expression of the jth column of the complete dictionary matrix has been given in step 2;

[0046] 2) Initialization: Index Set Vector Collection Residual r0 = y and signal target number P;

[0047] 3) Calculation Get the index number of the column;

[0048] 4) Update the index set

[0049] 5) Update vector set

[0050] 6) Calculate sparse vector Update residual

[0051] 7) Determine whether the iteration termination condition is met. If not, loop steps 3)-7);

[0052] 8) According to the column index number of P targets obtained Calculate rough estimates of m, n, and s corresponding to the target parameters

[0053] In step 4, in order to avoid the trade-off between parameter estimation accuracy and algorithm complexity of the 3D-OMP algorithm, the present invention establishes a smaller off-grid search interval according to the rough estimate of the target parameter obtained in step 3, and establishes an off-grid initial search interval for the pth target as follows:

[0054]

[0055] The search process solves the following problem:

[0056]

[0057] in in, represents the Kronecker product operator. This problem can be solved again using the 3D-OMP algorithm, but there is still a trade-off between parameter estimation accuracy and algorithm complexity due to meshing. Therefore, the method in step 5 is used to solve it.

[0058] In the step 5, in order to avoid grid division and realize high-precision parameter estimation at the same time, a three-dimensional golden section algorithm is proposed based on the 3D-OMP algorithm. According to the three-dimensional golden section ratio, the search interval is uniformly sampled eight times, and the function is calculated respectively. The value of , find the local maximum point, and update the grid search interval Γ p , repeat the iteration many times to find the maximum value of the global function, which is the desired parameter Further, according to the formula:

[0059]

[0060] Get the estimated results of the target range-speed-azimuth parameters Among them, c represents the propagation speed of radio waves in free space, f c Represents the carrier frequency.

[0061] The advantages of the present invention are as follows: The traditional grid-based 3D-OMP method is completely limited by grid division, and a trade-off between parameter estimation accuracy and algorithm complexity is required. The off-grid 3D-OGOMP method proposed on the basis of 3D-OMP combined with the three-dimensional golden section can avoid this defect well. In the rough estimation stage, only rough grid division is required, which can effectively reduce the computational complexity; in the fine estimation stage, the off-grid search interval is uniformly sampled in combination with the three-dimensional golden section, which reduces the running time and improves the computational efficiency. At the same time, it avoids the mismatch problem that may be caused by grid division and improves the target parameter estimation accuracy. The combination of the two can achieve high-precision and low-complexity target parameter estimation in the MIMO-OTFS radar communication integrated system, providing a reference for the application in the actual intelligent transportation field. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a diagram of the integrated application scenario of MIMO-OTFS radar communication for intelligent transportation systems;

[0063] Figure 2 It is a flow chart of the implementation of the method for estimating the target off-grid parameters of the MIMO-OTFS radar communication integrated system of the present invention;

[0064] Figure 3 It is the specific process of off-grid parameter estimation method combining the three-dimensional golden section idea;

[0065] Figure 4 It is a comparison chart of simulation results of the traditional 3D-OMP method and the 3D-OGOMP method proposed in the present invention when the real target is on the grid;

[0066] Figure 5 It is a comparison chart of simulation results of the traditional 3D-OMP method and the 3D-OGOMP method proposed in the present invention when the real target is not in the grid;

[0067] Figure 6 It is a curve diagram showing the change of the root mean square error (RMSE) between the target distance estimation value and the true value obtained by the 3D-OGOMP method proposed in the present invention with the signal-to-noise ratio (SNR) under different target numbers;

[0068] Figure 7 It is a curve diagram showing the change of the root mean square error (RMSE) between the target velocity estimation value and the true value obtained by the 3D-OGOMP method proposed in the present invention with the signal-to-noise ratio (SNR) under different target numbers;

[0069] Figure 8 It is a curve diagram showing the change of the root mean square error (RMSE) between the target azimuth estimation value and the true value obtained by the 3D-OGOMP method proposed in the present invention with the signal-to-noise ratio (SNR) under different target numbers;

[0070] Fig. 9 3D-OMP method and the 3D-OGOMP method proposed in the present invention respectively show the distance RMSE changing with SNR;

[0071] Fig.10 3D-OMP method and the 3D-OGOMP method proposed in the present invention respectively show the speed RMSE versus SNR variation curves;

[0072] Fig.11 3D-OMP method and the 3D-OGOMP method proposed in the present invention respectively show the change curve of azimuth RMSE with SNR. DETAILED DESCRIPTION

[0073] Figure 1 This is a diagram of the integrated application scenario of MIMO-OTFS radar communication in intelligent transportation systems, in which a single base station adopts a uniform linear array (ULA) composed of N t transmit antennas and N rThe receiving antennas are composed of half-wavelength spacing. In this model, the vehicle travels on a straight road section. Combined with the target parameter estimation algorithm, the intelligently connected base station can realize real-time perception and positioning of the LoS target vehicle.

[0074] The technical solution of the present invention is described in detail below in conjunction with specific implementation methods. Figure 2 As shown, the present invention provides a method for estimating target grid parameters of a MIMO-OTFS radar communication integrated system, which specifically includes the following steps:

[0075] Step 1: Construct the receiving signal model of the MIMO-OTFS radar communication integrated system; assume that on the base station side, the classic OTFS resource unit structure is composed of M subcarriers and N symbols. In OTFS modulation, the quadrature amplitude modulation (QAM) information symbol is regarded as placed in the DD domain plane, which is recorded as:

[0076]

[0077] Where △f and T represent the subcarrier spacing and symbol duration respectively, and satisfy △f = 1 / T. The DD domain signal is expressed in matrix form as Consider a system with N t transmit antennas and N r MIMO-OTFS radar communication integrated system with 100 receiving antennas, using precoding matrix After spatial multiplexing, the DD domain signal transmitted on the i-th antenna is:

[0078]

[0079] Where W represents the Hadamard product operator. According to the need of target parameter estimation, the precoding matrix is ​​designed using the omnidirectional transmission strategy. Divide the entire data frame into mutually exclusive and exhaustive subsets to achieve N t The orthogonality between the antennas is calculated. After the inverse symplectic finite Fourier transform (ISFFT), the TF domain signal transmitted on the i-th antenna is:

[0080]

[0081] in and are the M×M dimensional discrete Fourier transform DFT matrix and the N×N dimensional inverse discrete Fourier transform IDFT matrix, respectively. H Represents the conjugate transpose operator. Performing Heisenberg transform on the TF domain signal, the time domain signal transmitted on the i-th antenna is obtained as:

[0082]

[0083] Among them, g tx (t) is the waveform function of the transmitted signal. Finally, a cyclic prefix (CP) is inserted into the time domain signal of each data frame. The time domain signal can be expressed as:

[0084]

[0085] Where T cp represents the cyclic prefix duration, That is the transmission signal of the MIMO-OTFS radar communication integrated system;

[0086] Assuming that there are P targets in the line-of-sight path LoS of the perception environment, the MIMO channel including delay, Doppler frequency shift and azimuth parameters can be expressed as:

[0087]

[0088] where α p is the channel gain, and denote the transmitting direction vector and the receiving direction vector respectively, (·) T Represents the transpose operator. p , ν p ,φ p and θ p They represent the time delay, Doppler shift, DOD, and DOA parameter information of the pth target respectively. δ(·) represents the delta function. After transmission through the channel, the received time domain continuous signal is:

[0089]

[0090] in Indicated by The signal vector without CP is formed. Indicated by The signal vector with CP is formed. represents the additive white Gaussian noise AWGN vector that is independent of the signal;

[0091] exist m=0,1,…,MN-1 for the received time domain signal vector After sampling, the matrix form of the received signal is:

[0092]

[0093] in Indicates rounding up, Π MN is a cyclic shift matrix of dimension MN×MN, I Nrepresents the identity matrix of dimension N×N, Indicated by The transmit signal matrix is ​​composed of Represents the sampled additive white Gaussian noise (AWGN) matrix. Considering a single-base radar system, the magnitude of DOD and DOA are the same, so both are represented by the symbol θ p represents the azimuth, is a diagonal matrix describing the Doppler frequency shift characteristics of the target, expressed as:

[0094]

[0095] represents the diagonal matrix containing the target delay effect, expressed as:

[0096]

[0097] These two diagonal matrices retain the real delay and Doppler frequency shift effects in the channel matrix and reconstruct the real physical channel. In summary, the matrix That is, the receiving signal matrix of the MIMO-OTFS radar communication integrated system;

[0098] Step 2, vectorize and sparsely represent the received signal matrix, and express the target parameter estimation problem as a sparse signal recovery problem; first, in order to use compressed sensing theory to realize target parameter estimation, the received signal matrix Y is vectorized and expressed as:

[0099]

[0100] Among them, vec(·) represents the vectorization operator. Since the number of targets in the entire space is small, the following sparse signal representation model can be constructed:

[0101] y=Φα+z

[0102] in, is a dictionary matrix, is a sparse vector, is the noise vector, G t , G ν and G θ are the number of grid points corresponding to the three parameters of delay, Doppler frequency shift and azimuth respectively. The purpose of the above sparse signal recovery problem is to recover the sparse vector α by receiving the signal vector y and the dictionary matrix Φ. Here, the vector of the jth column of the dictionary matrix is ​​expressed as:

[0103]

[0104] where j∈[1,G t G ν Gθ ],j=mns,m=0,…,G t -1,n=0,…,G υ -1,s=0,…,G θ -1, therefore, if the values ​​of m, n and s are obtained, the formula and Calculate the corresponding time delay, Doppler frequency shift and azimuth values ​​of each column respectively;

[0105] Step 3: Use the grid-based 3D-OMP algorithm to roughly estimate the target parameters. Specifically, based on the sparse signal recovery problem in step 2, use the 3D-OMP algorithm to obtain rough estimates of the target delay, Doppler frequency shift, and azimuth. The specific process of the 3D-OMP algorithm is as follows:

[0106] 1) Constructing an overcomplete dictionary matrix Φ: The idea of ​​three-dimensional sparse signal recovery can be understood as searching on a three-dimensional grid set Γ, which can be expressed as:

[0107]

[0108] The vector expression of the jth column of the complete dictionary matrix has been given in step 2;

[0109] 2) Initialization: Index Set Vector Collection Residual r0 = y and signal target number P;

[0110] 3) Calculation Get the index number of the column;

[0111] 4) Update the index set

[0112] 5) Update vector set

[0113] 6) Calculate sparse vector Update residual

[0114] 7) Determine whether the iteration termination condition is met. If not, loop steps 3)-7);

[0115] 8) According to the column index number of P targets obtained Calculate rough estimates of m, n, and s corresponding to the target parameters

[0116] Step 4: Establish an off-grid search interval based on the rough estimate of the target parameter. Specifically, in order to avoid the trade-off between parameter estimation accuracy and algorithm complexity of the 3D-OMP algorithm, a smaller off-grid search interval is established based on the rough estimate of the target parameter obtained in step 3. The off-grid initial search interval for the pth target is established as:

[0117]

[0118] The search process solves the following problem:

[0119]

[0120] in in, represents the Kronecker product operator. This problem can be solved again using the 3D-OMP algorithm, but there is still a trade-off between parameter estimation accuracy and algorithm complexity due to mesh division. Therefore, the present invention adopts the method in step 5 to solve it;

[0121] Step 5: Combine the three-dimensional golden section idea to achieve accurate estimation of the target distance, speed and azimuth parameters from the grid. In order to avoid grid division and achieve high-precision parameter estimation at the same time, a three-dimensional golden section algorithm is proposed based on the 3D-OMP algorithm. According to the three-dimensional golden section ratio, the search interval is uniformly sampled eight times, and the function is calculated respectively. The value of , find the local maximum point, and update the grid search interval Γ p , repeat the iteration many times to find the maximum value of the global function, which is the desired parameter Further, according to the formula:

[0122]

[0123] Get the estimated results of the target range-speed-azimuth parameters Among them, c represents the propagation speed of radio waves in free space, f c represents the carrier frequency. The specific process of the off-grid parameter estimation method in this step is as follows Figure 3 shown.

[0124] The effect of the present invention is further verified by MATLAB simulation experiment below.

[0125] The simulation parameters are set as follows: the number of subcarriers in the OTFS frame structure of the radar communication integrated system is M = 64, the number of symbols is N = 16, and the carrier frequency is f c =77GHz, subcarrier spacing △f = 156.25KHz, total signal bandwidth B = 10MHz. Assume that the array element spacing d = λ2, the number of transmitting antennas N t=16, number of receiving antennas N r = 16. The number of distance grid points G in the simulation algorithm t =64, velocity grid points G ν =16, azimuth grid points G θ =36.

[0126] Figure 4 , 5 The figure is a comparison chart of simulation estimation results of the traditional 3D-OMP method and the 3D-OGOMP method proposed in the present invention, considering the following two situations: Figure 4 For the case where the real targets are on the grid, the real values ​​of the range-velocity-azimuth of the three targets are set to [74.95m, 57.03m / s, 50.00°], [134.91m, 19.01m / s, 85.00°], and [104.93m, 38.02m / s, 110.00°], respectively. Figure 5 For the case where the real target is not on the grid, the real parameter values ​​of the three targets are set to [70.89m, 61.62m / s, 47.75°], [138.91m, 23.12m / s, 83.50°] and [108.86m, 42.03m / s, 108.50°]. The above experiments were carried out under the condition of SNR=20dB, and the estimated value is the average of 100 Monte Carlo experiments. The results show that compared with the traditional grid-based 3D-OMP method, the off-grid 3D-OGOMP method proposed in the present invention breaks through the limitations of grid division and can still ensure high estimation accuracy when the real target is not on the grid.

[0127] Figure 6 , 7 8 is a graph showing the change of the root mean square error (RMSE) between the target parameter estimation value and the true value obtained by the 3D-OGOMP method proposed in the present invention with the signal-to-noise ratio (SNR) under different target numbers. Here, the case where the real target is not on the grid is considered, and the parameters and Figure 5 The parameters are the same. Figure 6 The distance comparison curve of the target is shown in Figure 2. Figure 7 The speed comparison curve of the target is shown in Figure 2. Figure 8 It is a comparison curve of the azimuth angle of the target. It can be seen that the RMSE between the estimated value and the true value decreases with the increase of SNR. Under the same SNR situation, the RMSE increases with the increase of the number of targets P. This is because the received signal is a superposition of multiple target echo signals, and the echo signals between each target will interfere with each other, resulting in a decrease in the estimation accuracy. However, since the method of the present invention adopts the off-grid parameter estimation technology, even in the case of three targets, a very small estimation error can still be guaranteed.

[0128] Fig. 9 , 10 11 is the RMSE versus SNR curve of the traditional 3D-OMP method and the 3D-OGOMP method of the present invention, and the actual parameter value of the target is set to [70.89m, 61.62m / s, 47.75°]. Fig. 9 It is the distance RMSE-SNR curve diagram, Fig.10 It is the speed RMSE-SNR curve diagram, Fig.11 It is a azimuth RMSE-SNR curve diagram. It can be seen from it that under different SNR conditions, the RMSE of the distance, speed and azimuth of the grid-free 3D-OGOMP method proposed by the present invention are all smaller than those of the traditional grid-based 3D-OMP method, indicating that the method of the present invention has higher estimation accuracy. In particular, as the SNR increases, the RMSE of the parameter estimation of the traditional 3D-OMP method will not be further reduced due to the limited grid division, but the RMSE of the 3D-OGOMP method of the present invention will become smaller and smaller. Therefore, high-precision target distance, speed and azimuth estimation can be achieved, which is more suitable for target vehicle perception in the field of intelligent transportation.

Claims

1. A method for estimating target grid parameters of a MIMO-OTFS radar communication integrated system, characterized in that: The following steps are involved: Step 1: construct a receiving signal model of the orthogonal time-frequency-space (OTFS) modulation multiple-input multiple-output (MIMO) radar communication integrated JRC system; Step 2, vectorize and sparsely represent the received signal matrix, and formulate the target parameter estimation problem as a sparse signal recovery problem; Step 3: Based on the sparse signal recovery problem in step 2, a grid-based three-dimensional orthogonal matching pursuit (3D-OMP) algorithm is used to roughly estimate the target parameters. Step 4, establishing smaller off-grid search intervals for multiple targets based on rough estimates of target parameters; Step 5, combining the three-dimensional golden section idea, uniformly sample the initial off-grid search interval eight times, calculate the function value, update the search interval according to the maximum value, and after multiple repeated iterations, achieve accurate estimation of the target distance-speed-azimuth.

2. A method for estimating target grid parameters of a MIMO-OTFS radar communication integrated system according to claim 1, characterized in that: In step 1, it is assumed that at the base station side, a classic OTFS resource unit structure is composed of M subcarriers and N symbols. In OTFS modulation, the orthogonal amplitude modulation QAM information symbol is regarded as being placed in the DD domain plane, which is recorded as: Where △f and T represent the subcarrier spacing and symbol duration respectively, and satisfy △f = 1 / T, the DD domain signal can be expressed in matrix form as Consider a system with N t transmit antennas and N r MIMO-OTFS radar communication integrated system with receiving antennas, using precoding matrix After spatial multiplexing, the DD domain signal transmitted on the i-th antenna is: Where W represents the Hadamard product operator. According to the need of target parameter estimation, the precoding matrix is ​​designed using the omnidirectional transmission strategy. Divide the entire data frame into mutually exclusive and exhaustive subsets to achieve N t The orthogonality between the antennas is determined by the symplectic inverse Fourier transform (ISFFT). The time-frequency TF domain signal transmitted on the i-th antenna is: in and are the M×M dimensional discrete Fourier transform DFT matrix and the N×N dimensional inverse discrete Fourier transform IDFT matrix, respectively. H represents the conjugate transpose operator, and the Heisenberg transform is performed on the TF domain signal to obtain the time domain signal transmitted on the i-th antenna: Among them, g tx (t) is the waveform function of the transmitted signal. Finally, a cyclic prefix CP is inserted into the time domain signal of each data frame. The time domain signal can be expressed as: Where T cp represents the cyclic prefix duration, That is the transmission signal of the MIMO-OTFS radar communication integrated system; Assuming that there are P targets in the line-of-sight path LoS of the perception environment, the MIMO channel including delay, Doppler frequency shift and azimuth parameters can be expressed as: where α p is the channel gain, and denote the transmitting direction vector and the receiving direction vector respectively, (·) T represents the transpose operator, t p , ν p ,φ p and θ p They represent the time delay, Doppler frequency shift, DOD, and DOA parameter information of the pth target respectively. δ(·) represents the delta function. After transmission through the channel, the received time domain continuous signal is: in Indicated by The signal vector without CP is formed. Indicated by The signal vector with CP is formed. represents the additive white Gaussian noise AWGN vector that is independent of the signal; exist m=0,1,…,MN-1 for the received time domain signal vector After sampling, the matrix form of the received signal is: in Indicates rounding up, Π MN is a cyclic shift matrix of dimension MN×MN, I N represents the identity matrix of dimension N×N, Indicated by The transmit signal matrix is ​​composed of Represents the sampled additive white Gaussian noise AWGN matrix. Considering a single-base radar system, the magnitude of DOD and DOA are the same, so both are represented by the symbol θ p represents the azimuth, is a diagonal matrix describing the Doppler frequency shift characteristics of the target, expressed as: represents the diagonal matrix containing the target delay effect, expressed as: These two diagonal matrices retain the real delay and Doppler frequency shift effects in the channel matrix and reconstruct the real physical channel. That is the receiving signal matrix of the MIMO-OTFS radar communication integrated system.

3. A method for estimating target grid parameters of a MIMO-OTFS radar communication integrated system according to claim 1, characterized in that: In step 2, first, in order to use the compressed sensing theory to achieve target parameter estimation, the received signal matrix Y is vectorized and expressed as: Among them, vec(·) represents the vectorization operator. Since the number of targets in the entire space is small, the following sparse signal representation model can be constructed: y=Φα+z in, is a dictionary matrix, is a sparse vector, is the noise vector, G t , G ν and G θ are the number of grid points corresponding to the three parameters of delay, Doppler frequency shift and azimuth respectively. The purpose of the above sparse signal recovery problem is to recover the sparse vector α by receiving the signal vector y and the dictionary matrix Φ. Here, the vector of the jth column of the dictionary matrix is ​​expressed as: where j∈[1,G t G ν G θ ],j=mns,m=0,…,G t -1,n=0,…,G υ -1,s=0,…,G θ -1, therefore, if the values ​​of m, n and s are obtained, the formula and Calculate the corresponding time delay, Doppler frequency shift and azimuth values ​​for each column respectively.

4. A method for estimating target grid parameters of a MIMO-OTFS radar communication integrated system according to claim 1, characterized in that: In step 3, based on the sparse signal recovery problem in step 2, the 3D-OMP algorithm is used to obtain a rough estimate of the target delay, Doppler frequency shift and azimuth. The specific process of the 3D-OMP algorithm is as follows: 1) Constructing an overcomplete dictionary matrix Φ: The idea of ​​three-dimensional sparse signal recovery can be understood as searching on a three-dimensional grid set Γ, which can be expressed as: The vector expression of the jth column of the complete dictionary matrix has been given in step 2; 2) Initialization: Index Set Vector Collection Residual r0 = y and signal target number P; 3) Calculation Get the index number of the column; 4) Update the index set 5) Update vector set 6) Calculate sparse vector Update residual 7) Determine whether the iteration termination condition is met. If not, loop steps 3)-7); 8) According to the column index number of P targets obtained Calculate rough estimates of m, n, and s corresponding to the target parameters 5. A method for estimating target grid parameters of a MIMO-OTFS radar communication integrated system according to claim 1, characterized in that: In step 4, in order to avoid the trade-off between parameter estimation accuracy and algorithm complexity of the 3D-OMP algorithm, a smaller off-grid search interval is established based on the rough estimate of the target parameter obtained in step 3. The off-grid initial search interval for the pth target is established as: The search process solves the following problem: in in, represents the Kronecker product operator. This problem can be solved again using the 3D-OMP algorithm, but there is still a trade-off between parameter estimation accuracy and algorithm complexity due to meshing. Therefore, the method in step 5 is used to solve it.

6. A method for estimating target grid parameters of a MIMO-OTFS radar communication integrated system according to claim 1, characterized in that: In step 5, in order to avoid grid division and achieve high-precision parameter estimation at the same time, a three-dimensional golden section algorithm is proposed based on the 3D-OMP algorithm. According to the three-dimensional golden section ratio, the search interval is uniformly sampled eight times, and the function is calculated respectively. The value of , find the local maximum point, and update the grid search interval Γ p , repeat the iteration many times to find the maximum value of the global function, which is the required parameter Further, according to the formula: Get the estimated results of the target range-speed-azimuth parameters Among them, c represents the propagation speed of radio waves in free space, f c Represents the carrier frequency.

Images